bug#59038: loading this base64 file makes emacs -Q 28.2 peg a core infinitely

[Alan Mackenzie]

Hello, Eli.

On Sun, Nov 06, 2022 at 15:52:33 +0200, Eli Zaretskii wrote:
> > Date: Sun, 06 Nov 2022 09:18:23 +0000
> > From: Gregory Heytings <gregory@heytings.org>
> > cc: Eli Zaretskii <eliz@gnu.org>, 
> >     Gerd Möllmann <gerd.moellmann@gmail.com>, 
> >     59038@debbugs.gnu.org, checker@d6.com, acm@muc.de


> > >> That file opens just fine in other modes

> > > It also opens fine in c-mode, with global-font-lock-mode disabled.


> > Indeed, obviously it's a c-mode font-locking related bug.

> > >> Note that this bug has nothing to do with long lines.

> > > I imagine that the font-lock issue is related to the line being in 
> > > excess of 21,000 chars (but general redisplay obviously doesn't have 
> > > problems with lines this 'small').

> > > Reduced to 10,208 chars that file opens instantly under emacs -Q in 
> > > c-mode with font-lock enabled; but at 10,209 chars it hangs Emacs (I 
> > > killed it after waiting 4 minutes).


> > Interesting, thanks for the bisection!

> It's an infloop in c-brace-stack-at.

> What happens is that c-brace-stack-at calls c-update-brace-stack with
> arguments: (nil 1) 8160 13160.  c-update-brace-stack then calls
> c-syntactic-re-search-forward, which finds nothing interesting and
> returns with point at 13160, but then c-update-brace-stack calls
> c-beginning-of-current-token, which returns point back to 8160.  And
> it goes on and on and on...

Thanks for the debugging.

> Alan, what can be done with this?

This:


diff -r 53717eda724c cc-engine.el
--- a/cc-engine.el      Sat Oct 29 09:42:47 2022 +0000
+++ b/cc-engine.el      Sun Nov 06 16:26:09 2022 +0000
@@ -6177,9 +6177,10 @@
            (setq s (cdr s))))
         ((c-keyword-member kwd-sym 'c-flat-decl-block-kwds)
          (push 0 s))))
-      ;; The failing `c-syntactic-re-search-forward' may have left us in the
-      ;; middle of a token, which might be a significant token.  Fix this!
-      (c-beginning-of-current-token)
+      (when (> prev-match-pos 1)      ; Has the search matched at least once?
+       ;; The failing `c-syntactic-re-search-forward' may have left us in the
+       ;; middle of a token, which might be a significant token.  Fix this!
+       (c-beginning-of-current-token))
       (cons (point)
            (cons bound-<> s)))))
 

-- 
Alan Mackenzie (Nuremberg, Germany).